CoreStory V3 Opens the Enterprise’s Most Critical Systems to AI Coding Agents
With CoreStory, leading coding agents from OpenAI and Anthropic correctly handled hidden business rules in 40% to 74%
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
![]()
CoreStory, the code intelligence platform for enterprise software development, today announced CoreStory V3, a new version of its platform that turns large and legacy codebases into a knowledge graph of how each system works, including its business rules and where each rule lives in the code. Engineers, analysts and AI coding agents work from that knowledge graph so their changes are grounded in how the system actually behaves. CoreStory is showing V3 at its booth at Microsoft Ignite in San Francisco, Nov. 17 to 20.
AI coding agents can read almost any code, but reading the code is not the same as knowing the rules it enforces. In the systems large enterprises actually run, often millions of lines of COBOL, RPG, PowerBuilder and .NET with little documentation and few people left who know them, those rules are rarely written down. An agent never told about them works without them, and most teams don’t realize their agents are operating without this context. The result is code changes that look correct, pass review, and break the business.
CoreStory isn’t a coding agent. It is the intelligence layer coding agents use to work more effectively inside an enterprise. CoreStory reverse-engineers a group of applications into a knowledge graph of their architecture, data models, workflows and business rules.
CoreStory V3 reads a codebase once and ties each business rule it finds to the exact lines that enforce it. It watches commits to the application and updates its knowledge graph as the code changes. It hands the relevant rules to a coding agent while the agent plans and edits, and flags any changes that conflict with a known rule. Any coding agent or harness that supports MCP can use it, including Claude Code, GitHub Copilot, Cursor and Codex.
Benchmark results
To evaluate V3, CoreStory tested the latest coding agents from OpenAI and Anthropic on change tickets in legacy codebases of up to 250,000 lines of COBOL, Java, and Python. Each ticket depended on a business rule that the ticket itself did not mention.
Working alone, agents fixed or flagged the missing rule in 4% to 36% of attempts, depending on the model. With CoreStory V3, the same agents improved to between 40% and 74%. When agents know the business rules, they draft changes that respect them.
These results extended beyond frontier models: a low-cost open-weight model using CoreStory V3 matched Claude Opus 5.5 using CoreStory at about one-tenth the cost per fix.
These gains held on everyday tickets as well: Claude Sonnet 5.5 kept 80% of unstated rules intact with CoreStory, against just 50% on its own. And CoreStory itself located 98% of the places where business rules live in a 250,000-line codebase, against 42% to 88% for frontier agents working alone.
Methods and full results: corestory.ai/research/v3-benchmarks.
“Agents keep extending their ability to build autonomously, but they make frequent, subtle mistakes in scope and understanding that keep them from working unsupervised,” said Anand Kulkarni, founder and CEO of CoreStory. “CoreStory gives them the understanding of real enterprise systems they need to execute accurately. That’s the difference between agents a team has to supervise day-to-day and agents a team can trust to make real changes.”
What teams use it for
- AI coding agent boost: give coding agents such as GitHub Copilot, Claude Code, Cursor, and Codex the business rules and system knowledge they need to change enterprise code correctly, on modernization projects and everyday tickets alike.
- Legacy code intelligence: map how a legacy system works, including business rules no one wrote down, before changing it.
- PRD and spec generation: produce requirements and specifications from the code as it actually runs.
- M&A technical due diligence: give deal teams a clear view of a target’s codebase and its risks before a deal closes.
Availability
CoreStory V3 is available starting Nov. 17, 2026. Teams can start with the free Developer plan at corestory.ai. Paid plans include Pro at $20 a month, Team at $80 a month, and Enterprise pricing is available on request.
About CoreStory
CoreStory is the code intelligence platform for enterprise software development. It builds a persistent understanding of how large and legacy codebases work and makes that understanding available to people and AI coding agents, everywhere product and engineering work happens. Learn more at corestory.ai.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261117234089/en/
Media gallery
